Research Of Fault Diagnosis Methods Based On Knowledge

Fault diagnosis is important to guarantee the quality and output of products in the process.With the development of the distributed system,a large amount of historical data can be obtained easily,which makes the fault diagnosis based on knowledge attented widely.However,with the emergence of various complex data,it leads to the fault diagnosis result unstatisfactorily.Therefore,it needs to be appropriate for data processing to get more classification characteristics and improve the diagnostic accuracy.So,this paper focuses on the fault diagnosis based on knowledge.The main research as follows:Firstly,summarized the fault diagnosis methods and the fault diagnosis based on knowledge methods.In this paper,we summarize the typica fault diagnosis algorithms of qualitative and quantitative fault diagnosis methods,and analyze the advantages and disadvantages of both,in addition,introduce the typical classification methods of fault diagnosis.Then,the Variable Weighted Joint Fisher Discriminant Analysis(VWJFDA)was proposed based on the statistical analysis method.It conbines the variable weighted and JFDA(Joint Fisher Discirminant Analysis)algorithm to improve the classification accuracy of diagnostic.Based on the idea,this paper also has carried on the experiment of VWLFDA(Variable-Weighted Local Fisher Discirminant Analysis).Next,with the idea of kernel,a Variable-Weighted Kernel Joint Fisher Discriminant Analysis(VWKJV)was proposed.The algorithm enlarges the fault characteristics by the method of nonlinear variable weighted,and combined with the strong ability of nonlinear data processing KJFDA(Kernel Joint Fisher Discriminant Analysis)algorithm for fault diagnosis.The proposed algorithm is verified effectively by experiment.Finally,the diagnosis method based on DBN and FDA(DBN-FDA)was proposed.This algorithm proposes another method to obtain the weighted variables vector.Namely,the variable weight vector can be obtained by tranning the deep belief networks.Then,based on the combined fault data,the fault diagnosis classification is made by combining with the FDA algorithm,and the comparison and analysis of the simulation experiments with other algorithms show the effectiveness of the DBN-FDA algorithm.