Fault diagnosis of blade crack based on multi-domain feature and information fusion
Aiming at the fault diagnosis problem of centrifugal fan blade crack,a blade crack fault feature ex-traction method based on multi-domain feature and information fusion was proposed.Firstly,on the basis of time-domain,frequency-domain and time-frequency-domain features,a series of cyclic-domain features were a-dopted for the amplitude-modulation characteristics of blade crack faults to construct a multi-domain feature set.Secondly,all the features in the multi-domain feature set were scored using various feature selection methods such as Laplacian score,random forest,ReliefF algorithm,mutual information,and information gain.Thirdly,the improved Dempster-Shafer evidence theory(DST)method was proposed to obtain a subset of sensitive fea-tures by fusing the feature score vectors under multiple criteria.Finally,a kernel principal component analysis(KPCA)method based on the optimized mayfly algorithm was proposed to make full use of the multi-sensor infor-mation.The extraction of sensitive features of the blade crack faults was completed and the diagnosis of blade crack faults was realized,The results show that the proposed method has an average testing accuracy of 99.70％,which is higher than those of other comparative methods,and is suitable for the fault diagnosis of blade crack.
blade crackfault diagnosiscyclic-domain featureinformation fusionDempster-Shafter evi-dence theory(DST)kernel principal component analysis(KPCA)
万方数据
维普
