Weighted Least Squares with Expectation-Maximization Algorithm for Burst Detection in U.K. Water Distribution Systems

2014 ◽  
Vol 140 (4) ◽  
pp. 417-424 ◽  
Author(s):  
Guoliang Ye ◽  
Richard Andrew Fenner
Keyword(s):  
Least Squares ◽  
Expectation Maximization ◽  
Distribution Systems ◽  
Water Distribution ◽  
Weighted Least Squares ◽  
Water Distribution Systems ◽  
Expectation Maximization Algorithm ◽  
Burst Detection

scholarly journals Optimal meter placement for pipe burst detection in water distribution systems

2016 ◽  
Vol 18 (4) ◽  
pp. 741-756 ◽  
Author(s):  
Medhanie Hagos ◽  
Donghwi Jung ◽  
Kevin E. Lansey
Keyword(s):  
Pipe Flow ◽  
Distribution Systems ◽  
Water Distribution ◽  
Water Distribution Systems ◽  
Roc Curves ◽  
Burst Detection ◽  
Flow Meters ◽  
Meter Placement ◽  
Pipe Burst ◽  
Optimal Meter Placement

Pipe bursts in water distribution systems (WDS) must be rapidly detected to minimize the loss of system functionality and recovery time. Pipe burst is the most common failure in WDS. It results in water loss out of the system, increased head losses, and low pressure at the customers' taps. Therefore, effective and efficient detection of pipe bursts can improve system resilience. To this end, this study proposes an optimal meter placement model to identify meter locations that maximize detection effectiveness for a given number of meters and type of meter. The linear programming model is demonstrated on a modified Austin EPANET hydraulic network. Receiver operating characteristic (ROC) curves for alternative pressure and flow meters are applied to investigate the relationship between the level of available information and pipe burst detection effectiveness. The optimal sensor locations were distinctly different depending on the type of meter and the objective to be considered. The ROC curves for alternative pressure and pipe flow meters showed that pipe flow meters are vulnerable to false alarms, and that using many pipe flow meters could detect all pipe bursts. Pressure meters could detect up to 82% of the burst events.

Sign in / Sign up

Export Citation Format

Share Document