A smart combination of observations and a computational model
Some leakages in water distribution networks are hard to find. But they show up in pressure and flow observations. Using a computational model of the network, it is possible to determine the location of a leakage from the observations.
Quickly demonstrated by combining Epanet and OpenDA
The idea is simple: build a computational model of the network and assume that a leakage of some unknown size is present in every point of the network. Then calibrate the model to make it match the available observations. Then every potential leakage should be zero, except for the one where the actual leakage is. A convenient tool for the calibration is OpenDA, an open source software toolbox with a variety of algorithms to combine observations and computational models.
By order of the Dutch water company Vitens, VORtech has demonstrated the validity of this idea. To do this, VORtech first coupled OpenDA to Epanet, which is well-known software for computations on distribution networks.
We have experimented with several OpenDA algorithms, but the initial results were not convincing. We then implemented a new algorithm in OpenDA which proved to work properly. A test was performed to see if the algorithm was capable of finding a leakage in an arbitrary location of the network. That was successful for 70% of the potential leakage locations.
In a follow-up development, we are now studying why certain leakages in certain locations are not properly found by the algorithm. We are also looking into several aspects that might confuse the algorithm. We are confident that we will be able to build an operational tool that can pinpoint leakages in distribution networks quickly and with sufficient accuracy,