Jonathan Jakob, André Artelt, Martina Hasenjäger, Barbara Hammer, "SAMknn Regressor for Online Learning in Water Distribution Networks", International Conference on Artificial Neural Networks (ICANN) 2022, 2022.

Water distribution networks are a key component of modern infrastructure for housing and industry. They transport and distribute water via widely branched networks from sources to the houses, buildings and industrial plants where it is consumed. In the flow of these networks anomalies can manifest themselves e.g. through leakages and or other unforeseen behaviour like fire runs. Since, each anomaly has the potential of being a leakage problem where large amounts of water simply get wasted, it is important to detect and localize them in order to prevent the waste. At the same time it is necessary to accurately predict the flow rate at certain nodes in the network to ensure that enough water is supplied. In this publication we use an adaption of the incremental SAMknn classifier for regression to build a leakage detection system that can adapt itself to the new flow rate after a leak occurs. Afterwards, when the leak is sealed it will automatically adapt back to the normal circumstances. We use data from simulated and real world water networks to showcase the usability of our approach.