Forecasting embankment dam behaviour with artificial intelligence

E. Masse (1) , C. Curt (1) , M. Le Goc (2)

1 UR OHAX - Ouvrages hydrauliques et hydrologie
2 LSIS - Laboratoire des Sciences de l'Information et des Systèmes

Forecasting dam behaviour is of paramount importance to avoid collapses. It consists of detecting and controlling deterioration mechanisms. This paper sets out to propose a multi model methodology to model complex dynamic systems, such as dams. It defines three models based on expert knowledge: a structural model, a functional model and a behavioural model. They are developed from the stories of the ageing of the processes that is to say from scenarios described in terms of events. These scenarios are interpreted by expert knowledge. One of the main ideas is to use the level of abstraction of the experts to facilitate the problem solving reasoning. A conceptual model which is created from CommonKADS methodology allows us to divide consistently the knowledge base into the three models. We illustrate this methodology by monitoring the dam behaviour during its life. Currently, the three models are developed from the scenario of the ageing of a French dam. They are validated by experts who gave their accordance with the developed approach. These models show that it is possible to formalize models implicitly used by experts.

Mots clés: BARRAGE; SYSTEME EXPERT; EXPERTISE; APPROCHE MULTI MODELE; SECURITE DES BARRAGES; VIEILLISSEMENT DES BARRAGES

Domaines: Intelligence artificielle [cs.AI]; Modélisation et simulation; Hydrologie; Milieux et Changements globaux; Sciences de l'environnement

E. Masse, C. Curt, M. Le Goc. Forecasting embankment dam behaviour with artificial intelligence. 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG), Oct 2008, Goa, India. 8 p. ⟨hal-00468567⟩