Research On Rolling Bearing Fault Diagnosis Under Varying Speed Conditions Based On Fractional Fourier Transform

The health condition of rolling bearings is essential to the rotating machinery operation.However,complex working conditions and harsh working environments will inevitably cause bearing faults.Therefore,it is necessary to perform efficient and accurate fault diagnosis on rolling bearings.This thesis focuses on the faulty bearing under varying speed conditions,and explores its vibration signal characteristics.Based on the fractional Fourier transform,the diagnosis methods are researched from the two perspectives of fault identification and classification.And the effectiveness of the researched methods is proved through simulation and experiment.The main contents are as follows:(1)Considering that the diagnosis method at constant speed is difficult to apply to varying speed bearings,the characteristics of the faulty bearing vibration signal at varying speed are explored.Aiming at two working conditions of constant speed and varying speed,the corresponding signal simulation models are constructed,and the bearing fault characteristics and vibration signal characteristics under respective working conditions are analyzed.The comparison shows the similarities and differences between the two,and the influence of varying speed on the signal.According to the vibration signal characteristics of varying speed faulty bearing,the advantages of the fractional Fourier transform under this condition are analyzed,and the basic theory is introduced.And the effect of the fractional Fourier transform on the simulated single-component chirp signal is explored.(2)A bearing fault identification method at varying speed based on optimized fractional filter is proposed.Refer to frequency band entropy,combined with short-time fractional Fourier transform,the fractional frequency band entropy is proposed to determine the location of the fault characteristic components in the corresponding fractional domain.In order to solve the problem of noise and multi-component interference,a global/local minimum fractional band entropy criterion is proposed to help determine the filtering center of the fractional filter.Through the above-mentioned ideas,multiple fractional filters are constructed to extract the fault characteristic components,and the envelope order spectrum is obtained by calculated order tracking to realize fault identification.Simulations and experiments prove that this method can effectively extract the fault characteristic components,and the comparison with other method also proves the advantages.(3)A bearing fault classification method at varying speed founded on multi-scale fractional dimensionless indicator is proposed.Based on the effective representation of the time-frequency domain by the fractional Fourier transform and the effective information mining by multi-scale analysis,combined with the advantage that the dimensionless indicator is not affected by working conditions,a new multi-scale fractional dimensionless indicator is proposed,to construct the original high-dimensional feature set.Combined with ReliefF algorithm for feature selection,and the selected sensitive feature subset is input into the random forest model for classification.Experimental signal analysis proves the excellent classification ability of this method.The comparisons of four different aspects further show the advantages of this method.