ECAI-2002 Conference Paper

[full paper]

Fault diagnosis of a continuous process using imprecise quantitative knowledge and causal models

Bruno Heim, Sylviane Gentil, Sylvie Cauvin, Louise Travé-Massuyes, Bertrand Braunschweig

Abstract. This paper presents a systematic methodology for building causal models that can be used for fault detection and isolation. The aim of a causal model is to capture the influences between the variables of a continuous process and to generate qualitative and quantitative knowledge that is interpreted by a diagnostic module. Following a model-based approach for fault detection, the diagnostic module compares the predicted outputs of the causal model with the measured values. Each influence of the causal model is associated with component(s) of the process. This qualitative knowledge is used to isolate the source fault on a set of components of the process. The application to a fluid catalytic cracking process pilot plant is briefly described and a fault scenario is finally presented. The work is done in the context of the EU-funded CHEM project "Advanced decision support system for Chemical/Petrochemical manufacturing processes".

Keywords: PAIS

Citation: Bruno Heim, Sylviane Gentil, Sylvie Cauvin, Louise Travé-Massuyes, Bertrand Braunschweig: Fault diagnosis of a continuous process using imprecise quantitative knowledge and causal models. In F. van Harmelen (ed.): ECAI2002, Proceedings of the 15th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2002, pp.648-652.