Fault Diagnosis Of Rolling Element Bearings Based On Optimal Demodulation Frequency Band Selection

As a critical component of rotating machinery,rolling element bearings have the capability of supporting a rotor and transforming sliding friction into rolling friction between the rotor and the stator.Thus it is widely utilized in various industrial occasions.Rolling element bearings have to operate continuously for long time duration and bear the dynamic load of the rotating machinery,which makes them subject to failure.The unexpected breakdown of the bearings will lead to serious performance degradation of the machinery as well as higher maintenance cost,or more seriously,some fatal accidents.Therefore,condition monitoring and diagnosis of rolling element bearings is of great significance to modern industry.However,bearings have to work together with other rotating components,such as gears and shafts.Therefore,the vibration signals collected from the rotating machinery are often mixed with a lot of noise.When the vibration signal generated by the bearing fault is weak,especially in the early stage of the fault,the existence of these disturbances from other rotating components will bring great difficulties in the diagnosis of bearing faults.To solve the problem of bearing fault diagnosis under strong interferences,it is of great importance to extract bearing fault features from complex vibration signals.In this thesis,rolling element bearings have been selected to be the research object,fault diagnosis of rolling element bearings under strong interferences,including cyclostationary noise and impulse noise caused by shaft,gear etc.is chosen to be the research target,the optimal frequency band for demodulation is set to be the main research threat.Two novel methods are proposed to select the optimal demodulation frequency band.The main research work and innovations of this thesis are summarized as follows:?1?The vibration signal model of faulty rolling element bearings is studied based on mechanical vibration.The disadvantages of analyzing the original spectrum are studied through the spectrum.?2?Based on the two classical methods for optimal demodulation frequency band selection–fast kurtogram?FK?and Protrugram,the procedure of frequency split and index calculation are thoroughly investigated.The principle and process of optimal frequency band selection are explained in details.?3?Kurtosis and square envelope spectrum are two core indicators used in classical demodulation frequency band selection methods.The relationship between these two indicators is explained through mathematical equations,and the reason why FK is heavily affected by cyclostationary noise is explained.A novel frequency band selection method(RCCgram_max)based on an improved cyclic component ratio is proposed to solve the problem that the performance of FK and Protrugram are affected by cyclostationary noise.At last,the influence of different levels of cyclostationary noise on the optimal frequency band selection is illustrated with simulation signals,and the superiority of RCCgram_maxax under cyclostationary noise is demonstrated.?4?After studying the generalized Gaussian cyclostationary model thoroughly,an optimal frequency band selection method?Discsgram?based on I_GGCS/GGS is proposed.The influence of different levels of impulse noise and cyclostationary noise on the optimal frequency band selection is studied with simulation signals and the superiority of Distcsgram under impulse noise and cyclostationary noise is demonstrated.?5?A novel envelope analysis procedure through combining RCCgram_max or Distcsgram with logarithmic square envelope spectrum is proposed.The proposed envelope analysis procedure is utilized to diagnose bearing faults under the circumstance of strong impulse noise/strong cyclostationary noise and planetary gear bearings,respectively.The results show that the proposed method is able to extract bearing fault features from complex vibration signals and diagnose bearing faults accurately.