| In complex modern industries,there are many devices and complex industrial processes in the production process,and industrial processes often need to be monitored,such as safety in industrial systems.At the same time,production safety is a crucial factor in industrial processes.This is the key cornerstone for the stable development of modern industry.Therefore,accurate positioning and traceability analysis are required for fault alarms in industrial processes.However,as the complexity of industrial processes increases,it becomes increasingly difficult to accurately locate and diagnose problems in their processes.Benefiting from the many advantages and characteristics of Bayesian network in the process of fault location and reasoning,it will greatly reduce the difficulty of fault diagnosis and location in complex industrial processes.However,the learning of the Bayesian network structure is particularly important.In the traditional Bayesian network structure learning algorithm represented by the K2 algorithm,this type of algorithm relies heavily on the node order in the node space,which makes the learned Bayesian network structure deviate from the real structure,so reasoning the results will also change accordingly.To solve this problem,this paper proposed an improved K2 algorithm,in this algorithm,compared with the traditional K2 algorithm,it can solve the problem that the input node order affects the learning result.That is,an initialized network structure is formed by the information entropy between the observable variables in the industrial process,the structure is topologically sorted,and a series of node order sequences are generated to improve the input of the K2 algorithm.In the algorithm,each node in the space is traversed greedily to generate a Bayesian network structure.Introducing information entropy to evaluate the structure can make the learned structure avoid overfitting.The learned network structure has better accuracy and better applicability.To realize the traceability analysis of the alarm,this paper combines the reasoning method of BN,and uses the posterior probability maximization as the criterion to carry out the traceability analysis.The method proposed in this paper is applied in the TE process,which indicated this proposed method can locate the source alarm variables effectively and accurately. |
