Data Classification Based On Maximum Likelihood Evidential Reasoning Rule And Its Application In Fault Diagnosis

As a more generalized extension of the Evidence Reasoning(ER)rule,Maximum Likelihood Evidential Reasoning(MAKER)rule makes up for the lack of considering the interdependence between multiple evidences in the process of ER fusion.For a piece of evidence abstracted from different attributes(features),the correlation factor is introduced to measure the common interdependence between multiple pieces of evidence,so as to improve the accuracy of multi-source information classification decision-making and create a more rigorous reasoning process for evidence fusion.Therefore,on the basis of fully considering the correlation between pieces of evidence,this dissertation designs a classifier based on MAKER rule to mine more correlation information for data classification.Then,based on the MAKER classifier,MAKER rule is applied to rotating machinery fault diagnosis under complete and incomplete samples.The main contents are as follows:(1)Design of generalized classifier based on maximum likelihood evidential reasoning rule.Firstly,principal component analysis(PCA)is used to select the features of the sample set.Then,based on the definition of correlation factor in MAKER rule,the degree of interdependence between evidences is obtained by constructing evidence space model(ESM),and then the MAKER rule is used to fuse the classification evidence,and the ER rule is used to realize the classification decision.Finally,experiments are carried out on the classic classification dataset provided by the University of California,Irvine(UCI)to verify that the performance of the MAKER classifier has a slight advantage over other classic classifiers.(2)Fault diagnosis method for rotating machinery based on MAKER rule.The generalized classifier in(1)is improved and MAKER rule is applied to fault diagnosis.In this method,the joint evidence space model(ESM)is constructed in the form of feature vector,and then the interdependence between multiple pieces of evidence is calculated.Then the MAKER rule is used to realize fusion and decision-making to predict the fault mode of test samples.In the fusion process,genetic algorithm is used to optimize the parameters of the diagnosis model.Finally,the effectiveness of MAKER diagnosis model is verified by motor rotor fault diagnosis experiment.(3)Fault diagnosis method of rotating machinery under incomplete samples conditions.Taking into account the fact that there is missing feature value data in actual fault diagnosis,a fault diagnosis method is further proposed based on the generalized classifier in(1).The lack of data is regarded as a natural characteristic of the sample characteristics,and participates in the fusion decision in the form of evidence.Finally,the rolling bearing fault diagnosis experiment verifies that the proposed method has an excellent performance in fault diagnosis under the condition of missing data.