Statistical models for assessing the technical state of water networks : some european experiences

This paper describes case studies where predictive service-life models are used for assessing the technical state of each individual pipe in a water distribution network. The goal is to enable planners to predict how the technical state of the pipes will change and, as a consequence, which are in most urgent need of repair. The mechanism of water pipes breaks is very complex and it is clearly difficult to create an applying physical model taking into account all the parameters. So statistical methods like lifetime models are emphasized. Historical failure records are then essential and together with network inventory data give the input for the analysis. The historical failure records are analyzed with a Weibull proportional hazard model. This paper describes some European practices in dealing with prediction of technical state of pipes. Data from two water utilities in France and from the city of Trondheim (Norway) are used in the analysis.