Indicateurs d'endommagement et durée de vie des réseaux d’assainissement, modélisation
Résumé
Building models of physical systems turns out to be an awkward task by the presence of noise and disturbance. It is not always possible to get information about disturbances and noises acting on the system. This may turn the usual stochastic framework inappropriate. In such cases, assuming bounded uncertainties may be a solution. An interesting way to go is then to use guaranteed estimation methods, which learn the state and/or parameters of the models from data. The set-membership models are well suited for fault detection and for making relevant diagnoses to improve the process evolution. A damage analysis of sewerage structures in Rennes (France) is undertaken. These systems (unit network rainwater and wastewater or separate system) consist of cement-based materials. The damage is quantified through various indicators of material properties (material extracted in the city). One hundred samples characterized by different streets and laying dates between 1892 and our days are analyzed. Thus, in this work, starting from a set-membership model of the behaviour of different sections of pipe wastewater, we use recent techniques of set-membership estimation to model and diagnose the system. A particular attention is paid to estimating the corrosion level for considering the full or partial rehabilitation.