Research On Bearing Fault Diagnosis Algorithm Based On The Improved Theories Of Frequency Band Entropy

As an important part of rotating machinery,rolling bearings play an important role in high-end CNC equipment,industrial robots and other mechanical equipment.Due to the influence of various complex factors,rolling bearings have become one of the components with the worst reliability in the whole rotating mechanical system.Their running state will have a direct impact on the running state of the whole mechanical equipment.The failure of rotating machinery is usually caused by the failure of rolling bearings.Therefore,relevant research is carried out for the condition monitoring and fault diagnosis of rolling bearings,which can evaluate the operating status of rotating equipment and provide a strong guarantee for the safe and stable operation of the equipment.At the same time the traditional regular maintenance or after-the-fact maintenance can also be transformed into condition-based maintenance(Equipment maintenance based on operating status),implement active maintenance of rolling bearings.Therefore,the condition monitoring and fault diagnosis of rolling bearings have always been hotspots and difficulties in the diagnosis of mechanical equipment.This academic dissertation is based on the theory of frequency band entropy method,with rolling bearings as the research object.The non-stationary vibration signal of the rolling bearing is analyzed and processed,and the fault characteristic information extraction of the vibration signal of the rolling bearing is deeply studied.According to the frequency band entropy method,the researches are carried out from two aspects: theory and application.It summarizes the shortcomings of the method due to the algorithm theory itself,and its limitations in the presence of strong noise interference,accidental impact,early weak faults and compound faults,and the corresponding solutions are proposed.The main work of this thesis are as follows:(1)The original frequency band entropy method is elaborated theoretically.The shortcomings of the original frequency band entropy method are analyzed and summarized:(1)The value of the bandwidth parameter of the original frequency band entropy method when designing the bandpass filter;(2)The defect of the timefrequency transform,that short-term Fourier transform(the window height and width of the STFT are fixed.It cannot take into account the needs of frequency resolution and time resolution,simultaneously);(3)The indicator used to determine the resonance frequency,that is,the information entropy is insufficient.(4)Constraint of the decomposition depth.At the same time,in various application scenarios of bearings,the frequency band entropy method also shows obvious limitations when there are strong noise,accidental impact interference,and compound failures.(2)According to the determination problem of the bandwidth parameter of the frequency band entropy method and the complexity of the rolling bearings working environment,which cause the introduction of strong background noise in the rolling bearing vibration signal,a solution combining the singular value decomposition preprocessing the original fault signal and the frequency band entropy method with bandwidth parameter optimized is proposed.First,on the premise of obtaining the resonance frequency,a bandwidth parameter optimization method based on the principle of kurtosis maximum is proposed,which can effectively improve the bandpass filter noise reduction performance of the frequency band entropy method.Secondly,to determine the reconstruction order of singular value decomposition,a model order determination method based on the relative change rate of singular kurtosis value is proposed to improve the passband noise reduction ability of singular value decomposition.This solution fully combines the excellent pass-band noise reduction ability of singular value decomposition and the excellent band-pass filter design ability of frequency band entropy to achieve complementary advantages,which can effectively weaken the interference of noise to the frequency band entropy method to find resonance frequency and realize fault feature extraction of rolling bearing under strong noise.In addition,the example analysis also verifies the effectiveness of the frequency band entropy method to achieve the optimal intrinsic mode function selection of the ensemble empirical mode decomposition.(3)In view of the inherent deficiencies of the frequency band entropy method,the lack of the time-frequency transform,indicator-information entropy,and the constraint of decomposition depth,as well as the collected rolling bearing vibration signal contains strong noise and may contain occasional impact interference,the performance of the frequency band entropy method is often limited.In this regard,this thesis has made an in-depth study and proposed corresponding solutions.(1)First of all,in view of the short-term Fourier transform's own shortcomings(the window height and width are fixed,and the requirements of frequency resolution and time resolution cannot be taken both into consideration at the same time),the introduction of wavelet packet transform into the frequency band entropy method is proposed to replace the short-term Fourier transform to get a better time-frequency distribution.Secondly,for the lack of index-entropy,the indicator of power amplitude spectrum entropy is proposed to determine the resonance frequency band.Finally,aiming at the constraint problem of decomposition level,a constraint method based on adaptive resonance bandwidth is proposed.Based on this,this thesis proposes an enhanced frequency band entropy method based on wavelet packet transform,power amplitude spectrum entropy and adaptive resonance bandwidth constraint.(2)Aiming at the deficiencies of enhanced frequency band entropy,a modified frequency band kurtosis method based on wavelet packet transform,envelope kurtosis index,cross-correlation coefficient and modified adaptive resonance bandwidth constraint is proposed to further overcome the deficiencies of the original frequency band entropy method and effectively overcome the enhanced frequency band entropy insufficient methods.(4)In practical engineering applications,the fault form of rolling bearings is often not single,but frequency band entropy and its improvement methods often cause missed diagnosis and misdiagnosis when diagnosing bearing composite faults.The optimized variational modal decomposition is introduced to preprocess the original fault signal,arrange the intrinsic mode functions in descending order and preliminary screening,and performing frequency band entropy analysis on the retained inherent mode functions in descending order.It can effectively solve the problem of bearing compound fault diagnosis and save workload.In addition,the experimental analysis verified the effective application of enhanced adaptive resonance technology based on power spectrum in the diagnosis of bearing compound faults.Both of the two methods can effectively realize the composite fault diagnosis of rolling bearings.