A Hybrid Method for Fault Detection and Modelling using Modal Intervals and ANFIS

Oftentimes the practical performance of analytical redundancy for fault detection and accommodation is decreased by the uncertainties associated to the model of the system and to the measurements. In this paper these uncertainties are taken into account through the definition of intervals for both the parameters of the model and the measurements. In the proposed method, a fault alarm is fired when an inconsistency between the behaviours of the system and the model emerges. Afterwards, the behaviour of the faulty system is modelled using an Adaptive Neuro Fuzzy Inference System (ANFIS). The identified model can be used for the fault accommodation task. The proposed method is applied to a simulated chemical plant. The obtained results highlight the capabilities for fault detection and accommodation of this method.