Research and Application of Weighted Fuzzy Unobservable Petri Nets Diagnosis Technology

Ruoming An,Minqiang Xu,Wei Liang

2012-01-01

Abstract:For complex system such as Hall thruster, the diagnosis problems are computation difficulty, unobservable event and unknown certain factor etc. For resolving these problems, a diagnosis model is proposed, which is based on weighted fuzzy Petri net with unobservable transitions. Fault class is defined by unobservable transitions subsets. The relationship between the set of unobservable transitions and the set of faults is established. Every fault class is departed into four levels of fault severity, which reflect the certainty value. Basis reachability graph is constructed as the foundation for diagnosis, which can be generated off-line when the Petri net is bounded. The set of minimal explanations is obtained by introducing two conceptions. The criterions to identify fault levels are derived by related theorem and constraint set. A two directional fuzzy reasoning strategy is adopted against observable or unobservable condition. The two directional reasoning are backward reasoning and forward reasoning. The level of alarm and certain factor is identified by an on-line diagnosis algorithm. The results applied to Hall thruster show that the diagnosis method is applicable and high efficient.