Bearing fault diagnosis based on SVMD and parameter optimized MCKD
Aiming at the problem that it was difficult to extract fault features from bearing fault signals with noise interference,a bearing fault diagnosis method combining the successive variational modal decomposition(SVMD)with the improved maximum correlation kurtosis deconvolution(MCKD)was proposed.Firstly,in order to characterize the faults in the bearing vibration signals,a weighted kurtosis index,which was more prominent than the kurtosis index,was proposed by combining the kurtosis with the Gaussian kernel.Secondly,the bearing signal was decomposed using the SVMD method to obtain a number of mode components,and the weighted kurtosis index was used to filter out the mode component with the richest fault features from the multiple mode components.Then,using envelope entropy as the standard,the geometric mean optimizer(GMO)was used to optimize the filter length and period of MCKD to obtain the optimal parameter combination.Finally,the GMO-MCKD method was used to reduce the noise of the bearing signals,and the noise-reduced signals were subjected to the envelope analysis to extract the bearing eigenfrequencies.At the same time,the variational mode decomposition(VMD)of particle swarm optimization(PSO)and variational mode extraction(VME)of particle swarm optimization were used for the comparison analysis of the bearing signals.The research results show that the SVMD-GMO-MCKD method diagnose a bearing characteristic frequency of 234.4 Hz and its second harmonic in the Cincinnati dataset,the bearing characteristic frequency diagnosed in the bearing dataset of Western Reserve University is 108.96 Hz and the second harmonic frequency is 218.09 Hz.The method can enhance the periodic shock component of the rolling bearing,effectively extract fault features of inner and outer rings of rolling bearings in the interference background,and the extraction effect of bearing fault features is superior to PSO-VMD and PSO-VME methods.
noise interferencesuccessive variational mode decomposition(SVMD)maximum correlated kurtosis deconvolution(MCKD)geometric mean optimizer(GMO)fault feature extraction effectbearing characteristic frequency
NETL
NSTL
万方数据
