The GERT Network For Fault Prediction Of Common Cause Failure Systems Based On Multi-Agent System

Due to the impact of common external factors,internal fault factors of several units in the system are stimulated at the same time,resulting in simultaneous failure,which is called common cause failure.Because of the statistical correlation between fault states,it is very difficult to analyze the reliability of the system.It is an important method of engineering design to increase the number of components and adopt redundancy method to improve the reliability of the system.But when there is common cause failure,the effect is not obvious by increasing the number of components to improve the reliability of the system.Therefore,under the condition of common cause failure,it is of practical significance to study the method of system reliability and availability analysis.In this paper,three key problems of common cause failure system fault prediction are studied,which are failure process description,failure parameter estimation and fault interval prediction.In the failure process description module,based on Poisson process,the occurrence pattern of common cause impact and the failure pattern of components in the system under impact are analyzed.According to the superposition of independent Poisson process,it is proved that the related parameters can be calculated with the observation results of multiple common cause failure.At the same time,this part also calculates the reliability function of components and systems under the background of common cause failure.In the parameter estimation module,based on the previous part,combined with the relevant theorems of probability theory and mathematical statistics,the expression of parameter estimation of common cause impact incidence and component failure probability is given.Based on the superposition of Poisson process,the data needed for parameter estimation are illustrated.In addition,Bayesian updating of parameters is realized by using new failure data.In the module of fault interval prediction,this paper overcomes the shortcomings of existing methods that are difficult to derive the reliability function of non-equal probabilistic common cause failure system,and constructs the failure path model of the system under common cause impact by using GERT network.At the same time,based on the traditional Mason's formula of signal flow graph,this paper studies the gain matrix of flow graph,expresses the relationship between nodes in GERT network by matrix,and calculates the equivalent transfer function of GERT network conveniently by using the operational property of matrix,so as to calculate the predicted value of system fault interval.In order to automatize and streamline the fault prediction and analysis of common cause failure system,three different levels of agents are established according to the mathematical and physical models of the above three modules,and their respective functional structures and cooperative working relationships are designed.By building a multi-agent system,the data information that needs to be input from outside for fault prediction can be identified,and the output results can be obtained quickly.Finally,an engineering example is given to show the application advantages of the multi-agent system.