| The healthy condition of machinery equipment is linked with the safe production and the reduction of cost of an enterprise.Bearing is one of the most widely used elements in machinery industry,thus making it very important to perform an accurate diagnosis of the bearing faults.The traditional cyclostationary analysis based on second order cyclic statistics is not effective in extracting the signature frequency of the fault under the non-Gaussian distributed background noise.In view of this,based on the Alpha stable distribution model,this paper is engaged in the study of a signal processing method based on the fractional lower order cyclic statistics.Revolving around the frequently damaged part—rolling element bearing,the simulated and real date signals of the outer race are used to prove the effectiveness of the method.The main research contents are concluded as follows:(1)Aim at the non-Gaussian distributed noise,the Alpha stable distributed model was used.Researching into the characteristics and the advantages and disadvantages of the commonly used fractional lower order statistics,combining the fractional lower order statistics with the traditional fault diagnostic methods,the spectrum analysis method based on fractional lower order statistics and envelope spectrum analysis method based on fractional lower order statistics were proposed.In order to prove the validity of the methods,simulations were conducted respectively.(2)The de-modulation principle of the fractional lower order cyclic statistics was theoretically deeply studied.The following simulations were performed: revolving around the amplitude modulated signal in Alpha stable distributed background noise,the second order cyclostationary analysis,the higher order cyclostationary analysis and the fractional lower order cyclostationary analysis were conducted respectively.The demodulation equivalence of the fractional lower order cyclostationary analysis was proven.With the change of α and SNR,it was found that the fractional lower order cyclic covariance spectrum and the phase fractional lower order cyclic covariance spectrum performed better than the cyclic covariation spectrum and the fractional lower order cyclic correlation spectrum in demodulation.(3)Bearing fault diagnosis method based on fractional lower order cyclic statistics was proposed.With regard to the simulated signal of bearing outer race fault in the non-Gaussian distributed background noiseand the real vibration data of the electric motor bearing outer race fault,the traditional cyclostationary analysis based on second order cyclic statistics and the fractional lower order cyclostationary analysis basedon fractional lower order statistics were conducted respectively.The outcomes verified the validity of the proposed method. |
