Research On Fault Diagnosis Methods Based On Petri Net

With the rapid development of modern industry and technology,it is important that the key equipment can operate in good state so as to improve the industrial economic benefits, even the national power. Therefore, it is necessary to research and develop various new effective fault diagnosis methods. Based on the character of fault diagnosis, a new method to solve the diagnostic problem is put forward, in which Petri net analytical method that can be good at describing the discrete and dynamic systems is applied in the fault diagnosis fields.On the point of application of Petri net in fault diagnosis fields, research work in the thesis focus on:1. Introduce the fault diagnosis methods briefly, in which knowledge-based fault diagnosis methods are emphasized, including fault diagnosis method based on Petri net, and discuss their merits and shortcoming in the application.2. Based on the basic fault character, a fault Petri net model is put forward by analyzing the application limits of traditional Petri net in the fault diagnosis fields. The model can well describe the character of fault evolution and development, as well as solve the problems of all kinds of fault development models. The diagnosis algorithm of fault Petri net is put forward, which has the ability to reason paralleling and to obtain all state results after reasoning, as well as quickly find out fault propagation routes. 3. On the base of the weighted fuzzy production rules and fuzzy Petri net, a weighted fuzzy Petri net model that fits the knowledge presentation and reasoning is brought forward, which can solve some problems such as the uncertain information and complex relations between fault and symptom in the fault diagnosis fields. A transition fired algorithm of weighted fuzzy Petri net is brought out by analyzing the character of the knowledge presentation and reasoning. At last, a backward-reasoning algorithm with the ability of retrospection is put forward and applied for knowledge reasoning.