A Method Of Behavior Analysis For Complex System Based On Hierarchical Bayesian Petri Net With Time Factor

Along with development of economy and technology, constantly improvement of industrial level and unceasingly expanding scale of system which is closely related to people’s life, the complexity and the difficulty of system modeling is also increase, which have an adverse effect on the maintenance of system security and stability. Faced with the complex system modeling and behavior analysis in various fields, Petri net is widely use because of it’s good description and analysis of the characteristics of iscrete event dynamic system behavior, but in practice, the number of nodes and state combinations of complex system increased with the increase of system complexity appears exponential growth, and is very easy to met the combination explosion when doing model analysis.So in the state explosion problem of large-scale complex system modeling analysis, this thesis puts forward a new kind of Petri net which based on the hierarchical Petri nets and the bayesian probability formula, the main work is as follows:(1) Proposed a model of Hierarchical Bayesian Petri Net With Time Factor(TF-HBPN), This model defines four mapping rules:class library map, transition map, and relationship map, or relationship map. This make modeling can be carried out in the process of hierarchical mapping, and the relationship between each layer model is more clear.(2) Proposed a recursive construction method based on this model,our method created top-level TF-HBPN for causal relationship of the observed system’s behavior, using the mapping rules for each level structure of the model upon the top-level model, and then handle concurrent behavior analysis of complex system problem decomposition with the hierarchical recursive ethod, transformed it into a simple causal relationship problem,(3) Proposed a recursive abductive behavior analysis method based on this model, get the causal chain of time by recursive abductive reasoning, analysis and compare to the known chronology, then calculate the occurrence probability of unobserved system behavior and the chain of events through Bayesian probability, finally, compared analysis result with normal action chain of events, isolate interference information.Instance analysis shows that the method can quickly modeling and analyze the large-scale complex system in many fields, it’s not affected by observed data interference and missing information while doing analysis of the system behavior and abductive reasoning of phenomenon and the credibility of analysis result is higher. Compared with other complex system analysis method based on Petri net, this method has lower modeling difficulty and model expression is more concise and easier to understand.