Application of a novel fuzzy classifier to fault detection and isolation of the DAMADICS benchmark problem

The performance of a novel fuzzy classifier when applied to fault detection and isolation of DAMADICS benchmark is investigated. The main properties of this methodology are the large accuracy with which it identifies the areas in the symptoms space corresponding to different categories, and the fine precision discrimination inside the overlapping areas. In the previous work one single category has been considered with the classifier for each one of the considered faults. Here, 20 levels of fault strength have been considered for each fault, ranging from small and often unnoticeable effects until large effects. The present work investigates the possibility to consider more than one category for each fault by considering different categories formed by single fault strengths or groups of fault strengths. This refinement offers a new insight and more information on the behavior of the faults, which improves isolation.