Research On Fault Feature Extraction Method For Rolling Bearing Under Variable Speed Condition

Rolling bearing is one of the most widely used mechanical parts in automobile,high-speed railway,aero-engine,numerical control machine tool,wind turbine and other rotating machinery.Its running state is very important for the safe operation of mechanical equipment.Due to the harsh and complex working environment,rolling bearing is a component with high failure rate,and it is more prone to failure in variable speed condition or start-up and stop stage.The traditional signal processing methods mainly study and analyze the signal under stable working condition,which is ineffective in variable speed signal processing.In addition,how to extract the fault feature from the test signal accurately is also a key problem to be solved when there are small or consistent faults in rolling bearing.Therefore,in the case of no phase acquisition equipment,this paper first studies the tacholess order analysis technology.Then,in the case of strong noise interference in the bearing vibration signal,the optimal frequency modulation band location method of bearing fault vibration signal is systematically studied to improve the accuracy of fault feature extraction of rolling bearing.The main achievements of this paper are as follows:(1)The tacholess order analysis technology based on vold Kalman is applied to the fault diagnosis of variable speed bearing.By combining vold Kalman adaptive filter(VKF)with time-frequency ridge extraction method based on penalty function,adaptive extraction of rolling bearing reference shaft harmonic component and related instantaneous phase information under the condition of tacholess signal is realized.The principle of parameter selection of VKF is determined.The envelope order spectrum is obtained by angular domain resampling and envelope analysis,and the conditional conversion of tacholess belief signal is realized Envelope order analysis of rolling bearing under speed condition.(2)In this paper,two improved spectral kurtosis methods are proposed for bearing vibration signal selection.Because the traditional spectral kurtosis(SK),squared envelope kurtosis(SESK)and cyclic component ratio(RCC)are sensitive to strong noise interference,the peak weighted harmonic kurtosis(PWHKS)is proposed by considering the influence of the amplitude of each harmonic frequency and the number of harmonics on the kurtosis and the impact cycle.Compared with SK and sesk,this index has stronger anti-interference ability to strong noise in vibration signal;by considering the influence of each harmonic frequency amplitude in RCC index of vibration signal,combined with the angle of energy,the composite weighted characteristic energy ratio(CWCER)is proposed.Compared with RCC index,this index has stronger anti-interference ability to strong noise in vibration signal.PWHKS and CWCER are frequency domain indexes which take into account impulse characteristics and cyclostationarity characteristics of impulse signals,and can accurately extract fault shocks from strong noise interference.Based on the two indexes mentioned above,PWHKSsgram and CWCERgram are proposed to locate the optimal de frequency band of bearing fault vibration signal accurately.(3)A fault feature extraction method based on tacholess order analysis and improved spectral kurtosis is proposed.Aiming at the vibration signal of variable speed fault bearing,firstly,VKF and ridge extraction method based on punishment function are used to obtain shaft instantaneous phase,and then angular domain resampling is used to obtain stationary vibration signal.Then,the optimal frequency modulation band of bearing is selected by using improved spectral kurtosis method(PWHKSgram,CWCERgram).Finally,envelope spectrum of band-pass filter signal is analyzed to obtain envelope order spectrum to realize variable rotation Speed bearing fault diagnosis.