Handling Modelling Uncertainty In Fault Detection And Isolation Systems

Abstract

The paper deals with the treatment of modelling uncertainties in model-based fault
detection and isolation (FDI). Essential for the practical implementation of model-based FDI
algorithms is to make accurate fault decisions despite the unavoidable deviations between the
model and the actual system under consideration. When analytical models are used, robustness of
the FDI algorithms is an important factor, which, however, is usually achieved on the cost of
increased complexity and often a reduction of the quality of FDI. A powerful alternative is the use
of qualitative models which allow accurate FDI under even imprecise observations and at reduced
complexity. In the first part we describe in some detail the basic concept of the analytical
approach - to lay the grounds - and in the second part we briefly outline the recent attempts to
employ non-analytical models while referring to the relevant literature for detail.