Detection and diagnosis of hybrid dynamic systems based on time fuzzy Petri nets

This paper presents a detection and diagnosis method based on a qualitative model of the process. Starting from an identification process a fuzzy partitioning of the variables evolution is made, defining for each measured variable a number of qualitative states. Then time fuzzy intervals representing the moment of state change are defined. The process behaviour is represented by time fuzzy Petri nets (TFPN). The evolution of the model is the consequence of events detection due to the partitioning bounds crossing. According to the membership possibility of an event to the estimated time interval it is possible to reason about a fault occurrence. The fuzzy data issue from the TFPN components allows evaluating the causes of the fault - the diagnosis.