Causal strategy for set-membership fault diagnosis

Set-membership fault detection is a promising approach because it takes into account a priori knowledge on modelling uncertainties and measurement errors. It consists in coherency tests with measurements and models. However, a careful diagnostic strategy has to be designed in order to achieve fault diagnosis of complex processes. To achieve the diagnostic procedure, a causal analysis allows focusing the coherency tests on simple models, resulting in a decrease of the amount of computations. This paper presents an application to a nuclear waste treatment plant. The proposed causal strategy for diagnosis is explained first. The set-membership coherency tests are then justified. Finally, a fault scenario is presented in order to enlighten the method.