Search

International audience; The safety control process of industrial systems (considered to be dynamic systems) needs to take in account physical processes (e.g. building ageing), informational processes (data collection and processing), decisional processes (data aggregation), and has to consider various constraints (e.g. economic and regulatory). The improvement of informational and decisional processes with the aim of controlling physical processes is based on the development of models and algorithms for measurement, assessment, control, diagnosis and prognostic. In the domain of dam management, assessment of reliability and safety, fault diagnosis, and corrective action proposals are carried out by expert engineers during dam reviews. With the perspective to assist these expert engineers, it is of great importance to develop methods and tools to manage the dynamic behaviour of dams and to model the processes at the same level of abstraction that is used by experts. In this chapter, the authors tackle the cognitive process of the diagnosis by means of a formal multi-modelling method and a diagnosis algorithm. The multi-modelling method called Timed Observations Method for Diagnosis (TOM4D) is based on the elaboration of four models: a Structural Model describing the relations between the components of the system, a Functional Model providing the relations between the values of the process variables (i.e. a set of mathematical functions), a Behavioural Model defining the states of the process and the discrete events firing the state transitions, and a Perception Model composed of a set of abstract variables, a set of thresholds associated to these variables and a set of constraints. The resulting process allows the automatic fault detection, identification and diagnosis and it is applied to hydraulic dam safety.

[Departement_IRSTEA]Eaux [TR1_IRSTEA]RIVAGE; International audience; This paper proposes the use of the Timed Observation theory as a powerful framework for model-based diagnosis. In fact, this theory provides a global formalism for modelling a dynamic system (TOM4D), for characterizing and computing diagnoses of the system under investigation.

[Departement_IRSTEA]Eaux [TR1_IRSTEA]RIVAGE; International audience; It is now well-known that the size of the model is the bottleneck when using model-based approaches to diagnose complex systems. To answer this problem, decompositional and multi modelling approaches have been proposed. In this paper, we propose a multi-modelling method called TOM4D (Timed Observations Modelling for Diagnosis) able to cope with dynamic aspects. It relies on four models: perception, structural, functional and behaviour models. The behaviour model is described through system component models as a set of component behaviour models and the global diagnosis is computed from the component diagnoses (also called local diagnoses). Another problem, which is far less considered, is the size of the diagnosis itself. However, it can also be huge enough, especially when dealing with dynamic system. To solve this problem, we propose in this paper to use The Timed Observation Theory. In this context, we characterize the diagnosis using TOM4D and the timed observation theory. We show their relevance to get a tractable representation of diagnosis. To illustrate the impact on the diagnosis size, experimental results on a hydraulic example are given.

[Departement_IRSTEA]Eaux [TR1_IRSTEA]RIVAGE; National audience; L'analyse et le contrôle du comportement des barrages en vue de maîtriser leur fiabilité et leur sécurité nécessitent la mise en ½uvre de modèles prenant en compte le passé (diagnostic), le présent (évaluation) et le futur de l'ouvrage (prédiction et proposition d'actions correctives). Nous avons montré au cours de travaux antérieurs qu'il était possible de modéliser le comportement de l'ouvrage par une approche multi-modèle qui rend possible la prise en compte de ces trois dimensions temporelles. Nous nous proposons dans cet article de montrer l'intérêt de cette modélisation pour l'évaluation des barrages en remblai et le diagnostic de leur comportement. / The analysis and control of dams' behaviour, with the aim of controlling reliability and safety, require the development of models that can take into account the present life of the dam (reliability and safety assessment), its past (diagnosis) and its future (prognosis and corrective actions). Previous works we carried out showed that it is possible to model the dam behaviour using a multi-model approach that takes into account the three temporal dimensions. In this paper, we propose to show the interest of such a modelling for the assessment and diagnosis of embankment dams.