Research On Fault Diagnosis Method Of Rolling Bearing Based On Vibration Signal Processing

As one of the critical components,rolling bearings are widely used in the rotating machineries.Their running states directly affect the performance of the mechanical equipments,even influence on the safety of the whole production lines.In order to ensure the smooth running of the equipments and reduce the occurrence of the accidents,it is of great significance to deeply carry out the research on the fault diagnosis and the state detection technology for rolling bearing.The rolling bearing is taken as the research object in this thesis.Focusing on the several key problems about the incipient fault diagnosis,the compound fault diagnosis,and the state recognition,the research work is conducted starting from the perspective of vibration signal analysis and processing.The main work and the obtained results of this thesis are as follows:The characteristics of variational mode decomposition(VMD)and empirical mode decomposition(EMD)are compared with each other through the simulated signals,and the influence of the penalty factor and the number of the components on the equivalent filter characteristic of VMD is qualitative analyzed.The composite interruption signal and the composite close-frequency signal are processed using VMD and EMD,respectively.The results show that VMD has more distinct advantages on restraining two kinds of mode aliasing problems.The equivalent filter characteristics of VMD and EMD are compared with each other by utilizing the numerical fractional gaussian noise simulation,and the results show that VMD has band-pass filter characteristic similar to wavelet packet transform.The qualitative influence of the penalty factor and the number of the components on the equivalent filter characteristic of VMD is analyzed and verified.It is shown that,the lager the values of the penalty factor and the number of components are,the narrower the passband of the equivalent filter bank will be.Otherwise,the affection result will be opposite.Facing on the problem that the incipient fault feature of rolling bearing is weak under the background noise environment and the incipient fault is difficult to identify,a weak fault diagnosis method based on the parameter optimization VMD is put forward.On the basic of analyzing the characteristics of particle swarm optimization algorithm,the influence of each control variable is discussed.Taking the advantages of the flexible signal processing characteristics of VMD,this algorithm is introduced into the bearing fault diagnosis.During the actual application,the final signal decomposition result is greatly affected by the selected penalty factor and the number of the components.In order to automatically determine the optimal parameter combination,the particle swarm optimization method is used to parallel search for these two influencing parameters.The analysis results,which are obtained from the simulated signal of the bearing incipient fault and the experimental signal of the whole life cycle accelerated fatigue indicate that,comparing with the conventional diagnosis methods,the proposed method is able to effectively identify the bearing weak fault and has obvious advantages.Facing on the problem that the compound fault features of rolling bearing are usually interfered with each other and accurate fault identification is difficult to achieve,a single channel compound fault diagnosis method based on maximum correlated kurtosis deconvolution(MCKD)and 1.5 dimension spectrum is proposed.The influences of the deconvolution period,the shift number,and the filter length on the MCKD signal processing results are discussed.In the process of the performing MCKD deconvolution,in order to acquire the optimal filter length,a two-phase grid searching method based on feature energy ratio index is utilized for the parameter selection.Taking the advantages of MCKD and 1.5 dimension spectrum,these two methods are respectively considered as the pre-processing method and the post-processing method to analyze the bearing fault signals.The processing results of the simulated and the tested compound fault signals demonstrate that,the proposed method can separate the faulty source signal corresponding to the single failure component from the original single channel signal,and accurate judgments on the bearing compound fault can be achieved.Taking into account the automation degree and the precision at the same time,a state recognition method combining the improved multiscale permutation entropy(IMPE),the dual tree complex wavelet packet transform(DTCWPT),the linear local tangent space alignment(LLTSA),and the extreme learning machine(ELM),is proposed.The IMPE theory is introduced into the bearing fault diagnosis.The influences of the chosen parameters on the entropy value calculation results are discussed,and meanwhile IMPE is compared with MPE through the simulated signals.Then IMPE is combined with the DTCWPT to depict the characteristics of the signal samples.In order to avoid the unfavorable effect of the information redundancy of the feature vectors on the recognition precision,LLTSA algorithm is utilized to reduce the dimensions of the feature vectors.In addition,K-fold cross validation method is used to select the neighborhood size and the intrinsic dimension.Taking the abilities of dealing with the small sample classification problem,ELM is regarded as the classifier to automatically classify the sample states.The verified results of the tested signal confirm that,the proposed method is able to identify the fault types and the fault degrees accurately and efficiently.The obtained results of this thesis provide new ideas for further research on the incipient fault diagnosis,the compound fault diagnosis,and the state recognition of rolling bearing.