Research On Fault Diagnosis Of Rolling Beairng Based On Non-Stationary Sequence Analysis

The rolling bearing vibration signal is coupled with vibrations of differentcomponents of the machine and the working environment is very complex, ithas the obvious of non-stationary. The key step is how to availably extract thefault features that can accurately reflect the state of the bearing from thevibration signals. Based on analysis of vibration signal of the rolling bearings,the focus of this article is the applying of empirical modedecomposition(EMD) and time-varying autoregressive(TVAR) models inrolling bearing fault diagnosis method.According to the blindness of selecting the IMF, a adaptive screeningidea of IMF based on the EMD energy threshold, is proposed. Then, theselected IMF is reconstructed, which is to lay the foundation for the furtherextraction.According to the problem of the traditional energy feature extraction,based on EMD, results in a low recognition rate of bearing fault, a new featureextraction method, extracting the energy ratio of IMF of the fault and normalsignal in the same conditions, is proposed. It is effectively for extracting theoperation characteristic and it provides a new idea for the on-line faultdiagnosis of bearing.According to the problem of the TVAR model is inaccurate in thetraditional feature extraction, a feature extraction method based on thedecision value of TVAR model order is proposed. Then, the problem is avoidby extracting the decision value of TVAR model order in the modeling processand the method provides a new feature extraction idea for bearing faultdiagnosis.Simulation experiments show that the proposed method can effectivelyand accurately identify the running state of rolling bearing.