The Research Of Alarm Root Cause Analysis Based On Bayesian Network

With the deepening of industrial information,alarm flooding has become one of the important safety problems in industrial process alarm management.The analysis of the root causes of the alarm plays a very important role in solving the alarm flooding problem.With the increasing complexity of the industrial process,the control algorithm is more and more diversified.How to effectively diagnose the root causes of the alarm is very important to solve the problem of alarm flooding.Bayesian network is a directed acyclic graph representing the causal relationship between variables based on probability theory and graph theory.It is used to solve the problem of uncertainty and probability.It has been widely used in artificial intelligence,machine learning,fault diagnosis and so on.But,in addition,if the number of nodes in the network becomes more and more in sometimes,it is very difficult to construct the Bayesian network through expert experience,and the solution space of the Bayesian network structure is exponentially increasing with the increase of the variables from data become an NP-hard research hotspot.The main research contents of this paper are as follows:(1)Introducing the current situation and research progress of alarm management system,expounds the research status and existing problems of Bayesian network,and studies the basic theory of Bayesian network.(2)Research on Bayesian network structure learning,through the application of cuckoo algorithm and simulated annealing algorithm The algorithm is used to prevent the algorithm from being locally optimized and the binary Levy strategy is used to update the structure and improve the accuracy of the algorithm.The algorithm is used to verify the learning efficiency of the algorithm.(3)The correlation between variables is determined by partial correlation analysis.Matrix,through the improved cuckoo algorithm,the TE process modeling and analysis,and through Bayesian reasoning on the diagnosis of the root causes of diagnosis and analysis,the experiment proved that the proposed algorithm in solving the uncertainty model construction and analysis of the feasibility and Effectiveness.