Passive Robust Fault Detection using Interval MA Parity Equations: Inverse vs Direct Image Tests

In this paper, a passive robust fault detection test based on calculating the inverse image of an interval model (linear or non-linear but linear with respect to the parameters) expressed in MA form is presented. This relies on a consistency test which uses tools from interval analysis and zonotope arithmetic to check if there exists a member in the family of models described by an interval model that can explain the measured data. The proposed test is compared to the classical robust interval model fault detection approach based on the direct image. The main features of both tests will be extracted and advantages and drawbacks discussed using a motivational example. Finally, an application example based on the known quadruple-tank process is used to assess how the algorithms work.