文摘
The presented study features a decision support system, alerting for contamination events in water distribution system (WDS). The developed model consist two modular elements: minimum volume ellipsoid (MVE) detecting outlier's measurements, and a following sequence analysis classifying contamination events. This study performs multi-dimensional analysis of the data that differs from the parallel one-dimensional analysis conducted so far. The application of an unsupervised classification method in the model eliminates the use of unfounded simulated events for the classifier construction and contributes its reliability and generality. The model was applied on a real WDS data, and showed high accuracy and detection ability.