A mixed-integer formulation for fault detection and diagnosis: Modelling and an illustrative example

In this paper, we propose a mixed integer optimization for diagnosis of fault events which changes the structure of the system model. The proposed approach aims at selecting the right model of the systems among a bank of models when some faults occur. This approach suits to the case where both abrupt faults (intermittent or permanent such as saturation or sudden shutdown) and incipient faults (continuous) are considered. The optimization helps finding the best combination of faults that have occurred. So doing, we go Further than analyzing incidence matrix of residual. Also, this approach allows to introduces fault occurrence logics such as the ones encountered when establishing fault trees.