Research On Equipment Fault Diagnosis Based On Full Vector EEMD And Case-based Reasoning

In modern manufacturing,equipment fault diagnosis technology,which can not only help enterprises find equipment failures in time and avoid major safety accidents,but also bring considerable economic benefits to enterprises,is receiving more and more attention.Therefore,equipment fault diagnosis technology has great research value.At present,most equipment fault diagnosis methods are based on knowledge rules,and it developed well,however,these diagnosis methods still have disadvantages that are difficult to overcome,such as difficulty in extracting knowledge rules and maintenance of knowledge rule bases.Aiming at the shortcomings of this method,this paper uses fault diagnosis method of the Case-Based Reasoning(CBR),which has the characteristics of getting knowledge easily,better growth and solving process simply,But constructing a fault diagnosis system of CBR,some problems are still encountered: there are some problems of losing fault information easily with the case library based on single channel information;it is difficult to process the nonlinear and non-stationary vibration signals for conventional signal analysis methods;the representational characteristics of cases are redundant;the conventional K-Nearest Neighbor(KNN)algorithm in case retrieval technology has low diagnostic accuracy rate.Based on the above problems,combined with the advantages of Full Vector Spectrum,Ensemble Empirical Mode Decomposition(EEMD)and CBR,a fault diagnosis method is proposed for rotating machinery based on Full Vector EEMD and CBR.The main research contents of this paper are divided into the following aspects:(1)Aiming at the problem that it is difficult to deal with the non-linear and nonstationary signals for conventional methods,and single channel signal is prone to miss fault information,the Full Vector Spectrum technology combined with EEMD method is used to separate noise signals and reduce mode mixing,and guarantee the integrity and accuracy of the fault information collected.Finally,experiments show that the fault features extracted based on the Full Vector EEMD method have better fault resolution capabilities.(2)Aiming at the problem of attribute redundancy in the CBR and the problem of low diagnosis accuracy rate of the KNN method,a optimization method of fault feature of rolling bearing combining Genetic Algorithm(GA)and Correlation Feature Selection(CFS)is proposed to screen the original feature set,reducing the redundancy between the features of the original feature set and significantly improve the calculation efficiency,but the diagnostic accuracy rate of the feature set optimized is slightly lower than the original feature set.In order to improve the accuracy of diagnosis,a optimization method of fault features of rolling bearing of GA-weighted KNN is proposed to add weight coefficients to the feature subset optimized,and effectively improving the diagnostic accuracy rate.Experimental results show that this method can significantly improve the diagnostic accuracy rate and calculation efficiency of rolling bearing fault diagnosis.(3)Based on Full Vector Spectrum,EEMD and CBR,the fault diagnosis model of rolling bearings is constructed.Relied on this model,mycbr modeling platform of CBR is used to build the fault diagnosis system of rolling bearings.Experimental results show that the system can diagnose not only a single fault type of rolling bearings,but also a complex fault type.Compared with a fault diagnosis system based on knowledge rules,the system does not need to study its complicated fault mechanism,the case is relatively easy to obtain,and the diagnosis results and solutions obtained are easier to understand and accept.