Research On Fault Diagnosis Method Of Chemical Process Based On Bayesian Network

With the development of statistics and machine learning,Bayesian network become more and more popular and has been applied in more and more areas.In this paper,the chemical TE process is analyzed,the analytic hierarchy process and the Bayesian network are used to model the complex chemical process.The theory and internal mechanism of complex and difficult to understand are displayed in an intuitive and visible network.At the same time,the fault diagnosis is achieved.In contrast to the traditional Bayesian network learning method,the Bayesian construction method based on AHP-K2 is a Bayesian network learning algorithm based on pure data,which overcomes the shortcomings that need to rely heavily on the experience of traditional experts.The Bayesian network in the field of fault diagnosis and application,and the traditional multi-variable statistical method-based data-driven technology in the application of fault diagnosis is different from the Bayesian network,the latter does not require strict assumptions on the Gaussian distribution,creating a new Mold method and its application in the field of chemical production,so it has important practical significance and research prospects.The process of the preparation consists of:(1)We study the Bayesian network theory,and propose a Bayesian network learning method based on AHP and K2 algorithm.The TE process is based on AHP-K2 Bayesian Network.(2)A Bayesian network hybrid structure learning algorithm based on constraint and scoring search is proposed.A Bayesian network structure learning algorithm for MI-AHP-K2 is proposed and compared with the empirical knowledge-based K2 algorithm on the Asia network.The experimental results show that the method is feasible and effective for the study of the network.(3)The AHP-K2 algorithm is applied to chemical fault diagnosis.The Bayesian network model is generated by using the analytic hierarchy process(AHP)to generate a Bayesian network model.It is used to analyze whether the system's operating state is changed by analyzing the fault analysis index such as aggregation coefficient,average degree and average distance.The fault condition is related to the normal operation time Network characteristics contrast,find the point of failure.The TE process model is studied,and the method proposed in this paper is verified by the TE process.It is proved that the method is feasible and effective.At present,the use of Bayesian theory in the process of fault diagnosis technology still has a large space for development,but also continue to experiment,while the strong theoretical support optimization algorithm is also a general direction.