Machine Learning Methods for Monitoring of Complex Water and Sewer Network Infrastructure

The aim of this project is to develop data-driven sensor placement and state estimation methods that predict water system problems with theoretical guarantees of robustness and reliability. The project is structured along the following objectives: 1. Develop optimal sensor placement strategies that maximize the amount of

information acquired by the sensing infrastructure. 2. Design and validate data-driven state estimation and fault detection algorithms that augment the information from the sensor network to the hydrodynamic models. 3. Design and validate machine learning forecasting tools that predict the state of the

network and key performance metrics, such as the occurrence of fault events, blockages, overflows, leaks etc.