Research On Condition Degeneration Of Rolling Bearing Based On Full Vector MEMD

Multivariate empirical mode decomposition(MEMD)and its extended algorithm named noise-assisted multivariate empirical mode decomposition(NA-MEMD),they collectively called MEMD in the paper,can decompose multi-channel signals synchronously,which provide a new way to study the rolling bearing's degradation condition in the same frequency scale combing with full vector fusion technique.The study object is vibration signals coming from the rolling bearing's full life-cycle.We study the change of frequency structure in the rolling bearing's degeneration process qualitatively,indicator selecting in that process,and remaining useful life of the bearing quantitatively.The main work is as follows:(1)In order to recognize the degradation process of rolling bearing,a method is proposed that combined MEMD with full vector fusion.Decomposing the multi-channel vibration signals which represent different working condition,a series of multivariate IMFs are obtained.The most sensitive IMF,which replacing the raw signals,is then selected by cross-correlation coefficient criterion to do the full vector envelope analysis for the sake of extracting the signal features.Analysis with simulated signal and actual signal provided by University of Cincinnati is done to test the effectiveness of the method respectively.It shows the regularity that the more serious of the fault condition is,the more complex of frequency spectrum structure becomes.A new train of thought is proposed to study the frequency spectrum structure of all the degradation process further.(2)Aiming at how to select and construct proper indicator of rolling bearings degradation process,a method combining NA-MEMD with principle component analysis is proposed to extract the degradation indicator.NA-MEMD is firstly used to decompose the two initial channels and a noise channel signals which stemming from full life-cycle,the sensitive IMFs are used to reconstructed raw signals,which are chosen by correlation coefficient criteria.After figuring out degradation indicator sequences of reconstructed signals in the degradation process,some good indicators are then extracted based on monotonicity and robustness of the sequences to do the PCA fusion.At last,the first principle component is considered as final indicator of rolling bearing's degradation process.The vibration signals of the rolling bearing are analyzed provided by PRONOSITS platform.It indicates that the indicator based on NA-MEMD and PCA represents the rolling bearing's degradation process better than any signal indicator.(3)A method combining the crack propagation model and particle filter is applied to predict the remaining useful life of the rolling bearing.It need to construct an indicator which can represent the rolling bearing's degeneration process,select Paris-Erdogan model as the degeneration model,designate the number of particles and threshold,and combine particle filter to predict the remaining useful life.The first principle component indicator is used to predict the remaining useful life based on NA-MEMD and PCA fusion.It shows the method has certain feasibility in the rolling bearing's remaining useful life prediction.