Risk analysis and diagnosis modelling for online control of process
Résumé
Currently safety systems are designed to warn operators when a critical fault happens. Unfortunately, this information may come too late and operators are then not able to correct the potential failures. In order to prevent system from critical faults and failure, this paper proposes a method to predict the risks based on a logical diagnosis approach and a prognosis of the consequences of the detected faults. Using the DX di-agnosis approach, some information are added so that physically impossible diagnoses are avoided and the faults differently isolated. Then, after an FMEA analysis of the system and its representation with the fault tree method, the consequences of the faults detected are propagated into the fault tree so that risks are prog-nosticated. Finally, once these risks determined, a second analysis of the fault tree gives the elements of the system to monitor in order to avoid, if possible, the previous risks.