Research On A Series Of Fault Feature Extraction Methods And Applications For Rotating Machinery

As the widely used production machinery,rotating machinery plays an important role in petrochemical,metallurgical,railway and other industries.Hence,it is of great practical significance to study the fault feature extraction technology of rotating machinery to ensure its reliability.Due to the abominable working conditions of the rotating machinery and the interrelation between the internal components,the fault features are easily to be submerged.Therefore,how to extract fault features accurately and effectively under heavy noise background is the unremitting pursuit of the research in the field of rotating machinery fault diagnosis.With the rotating machinery as the subject investigated,the research of its fault feature extraction methods is carried out in this paper.Under the background of heavy noise,the signals collected from rotating machinery usually exhibit nonlinear and non-stationary,and deviate from the real value,which creates difficulties for subsequent analysis.Aiming at this problem,a one-level kernel regression residual decomposition(KRRD)enhanced synchroextracting transform(SET)method is proposed.The method employs KRRD to extract the high frequency signal which contains the faulty information,SET is further used to the high frequency signal to denoise and the faulty feature frequency is demodulated through Hilbert envelope analysis.Aiming at the problems in the field of fault diagnosis for traditionaltime-frequency analysis(TFA)methods such as empirical mode decomposition(EMD),local mean decomposition(LMD)and intrinsic time-scale decomposition(ITD),i.e.mode mixing,end effects,signal mutation and signal distortion et al.,a kernel regression residual signal based improved intrinsic time-scale decomposition(IITD)method is proposed.By applying the proposed method to the fault bearings and gears,the results indicate that the proposed method overcomes the disadvantages of traditional TFA methods while achieving the satisfactory performance.Aiming at the expertise dependence on selection of the decomposed component in fault diagnosis,the kurtosis indicator(KI)is introduced and a method based on variational mode decomposition(VMD),KI and minimum entropy deconvolution(MED)is proposed.The method employs KI to select the optimal IMF which contains the faulty information,MED is further used to enhance the periodic impact characteristic.The proposed method achieves reliable identification of rotating machinery faults successfully.Aiming at the disadvantage of fault feature extraction of VMD,an autoregressive(AR)model enhanced VMD method is proposed.VMD is employed to decompose the signal into a set of intrinsic mode functions(IMFs),AR model is used to further purify each IMF.Through the engineering applications,the performance of the proposed method is verified.Compared with traditional VMD,AR model enhanced VMD has higher accuracy.