Research On Compound Faults Feature Extraction Method Of Rolling Bearings

Rolling bearings are one of the most important parts in rotating machinery.Different types of bearing failures can cause serious damage to the mechanical system.Long-running bearings are often accompanied by compound faults state that coexisting multiple faults.Compared with single fault,compound faults become difficult to extract their features due to mutual interference between multiple faults response.Aiming at this problem,based on the study of the vibration characteristics of rolling bearing compound faults,this paper proposed method to accurately extract the features of compound faults,starting from the situation that multiple faults can excite resonance frequency bands.The main research contents are as follows:1.Vibration characteristics of bearings with local faults are analyzed and summarized,and the case where the faults can excite the resonance frequency bands when the damage occurs are discussed.When compound faults occurred in a bearing,considering that the different bearing faults may provoke different resonant frequencies,and the resonance response of a same component can be excited by different local faults,a compound faults simulation model with partially overlapping multiple resonance bands is established,which makes the simulation more universal.Experimental analysis verified the correctness of the simulation.2.Aiming at the problem that kurtosis is susceptible to high peak pulse interference and resonance frequency bands with different energy,a method based on correlated kurtosis resonance demodulation is proposed,which highlights the periodic characteristics of bearing fault signal,and can effectively extract the respective dominant resonant frequency bands provoked by different faults.At the same time,considering that1.5-dimensional spectrum has the characteristic of eliminating frequencies that do not satisfy the non-linear coupling relationship.A compounds fault feature extraction method based on correlated kurtosis resonance demodulation and 1.5-dimensional spectrum is proposed,which realizes the separation of compound faults feature to some extent.3.Aiming at the overlapping and crossing phenomenon of resonance frequency bands caused by different faults,and the problem that the parameters of filter length and shift number in maximum correlated kurtosis deconvolution(MCKD)need to be selected by human experience,an adaptive maximum correlated kurtosis deconvolution(AMCKD)method based on artificial fish swarm(AFSA)is proposed and applied to compound faultsfeature extraction.Simulation,experiment and wind turbine gearbox bearing compound faults signal prove that the method can successfully decouple and accurately extract multiple resonance frequency bands of different bearing faults,and achieve accurate separation of compound faults feature of rolling bearing.