Research On The Nonstationary Vibration Signal Of Key Parts Of Gearbox And Its Fault Diagnosis Method

The vibration signals collected from gearbox surface perform the nonstationary and nonlinear properties due to the failure of key parts of gearbox and variation of its working condition.Additionally,the difference of dynamic feature of key components also makes the vibration signal becoming more complex.Hence,it is necessary to develop the advanced non-stationary signal analysis method to extract the effective equipment state information and complete the fault diagnosis.To meet the fault diagnosis demand of the gearbox key components(bearing and gear)we have conducted the in-depth research on non-stationary signal analysis and diagnosis technology to identify the bearing weak fault information,extract the gear damage feature,and monitor the health condition of gearbox under varying speed without tachometer reference in the paper.The contributions of the thesis are summarized as follows:(1)A new empirical wavelet transform(EWT)method is introduced into the weak bearing fault diagnosis to overcome the deficiency of empirical mode decomposition(EMD)and its similar method,which easily mistake the band location of bearing fault information.And then,a more suitable band detection method is selected by analyzing the basic property of EWT and the characteristic of faulty bearing signal.As a result,a new bearing fault diagnosis method is constructed based on the improved EWT method and the idea of suppressing noise in resonance frequency band.Finally,the simulation signal and two experimental cases verify the effectiveness of the proposed method and the correctness of the corresponding theoretical analysis.(2)Considering that the faulty bearing signal may contain multiple resonance frequency bands,we introduce the variational mode decomposition(VMD)method to improve the comprehensiveness of bearing fault diagnosis method.And then,a detailed discussion of the relationship between VMD and faulty bearing signal is given.Moreover,the parameter characteristic in VMD method is analyzed.Consequently,the multi-resonant component identification method for bearing fault diagnosis is proposed based on the VMD and the multi-resolution Teager energy operator(MTEO),which has a good ability for the purification.The simulation and test results show that the proposed method has the good performance for the bearing fault diagnosis,and outperform some common methods.(3)A nonlinear signal processing method named Detrend Fluctuation Analysis(DFA)is discussed to solve the problem of feature extraction of gear with similar damage.Firstly,we give the reason of crossover phenomenon buried in the analyzed result of DFA from the view of the frequency distribution,which are caused by the nonlinear mapping of DFA.Subsequently,based on the explanation of crossover phenomena,a set of feature with the specific physical meaning is put forward to character the gear damage degree.Finally,the experimental data verify that the different combinations of these proposed features can well deliver the identification of the gear damage degree,and the sensitive indices also show that some of the proposed features have the better sensitivity than some existed indicators.(4)In view of complex shape of curve drawn by combining the scale and fluctuation function of DFA method,two simplified scale methods are studied based on the analyzed result of crossover phenomenon,respectively.One is the simplified scale exponent identification method,which includes a local optimal scale identification method.Another one is the simplified curve shape method,which have a preprocess technique named the increment of extreme value.As a result,we establish two kinds of feature extraction methods based on the above introduced two simplified scale techniques.The analysis results of gearbox experimental data show that the proposed methods can well perform the gearbox condition monitoring under different situations.(5)The basic characteristic of ridge estimation based on cost function is analyzed.Then,a fusion ridge estimation method is proposed based on the features of the fine ridge with smoothness and failure ridge with break to realize the tacholess gearbox fault diagnosis under large speed variation.Furthermore,the basic fault diagnosis framework of gearbox under varying speed condition is established.As a consequence,the proposed ridge estimation method is verified by two groups of simulation signals and gearbox experimental data.And the characteristic frequency of a planet bearing with the inner race local defect is detected from the order spectrum based on the derived and discussed spectral structure of faulty planet bearing.(6)The out of work reasons of ridge estimation method based on cost function is firstly discussed to enhance its ability thoroughly to track the complex shape and weak ridge.And a new framework about the ridge estimation and a new local cost function with physical meaning are constructed,respectively.Hence,a path optimization ridge estimation method is proposed based on these enhanced measurements.The simulation signals with different signal noise ratio(SNR),the experimental gearbox vibration signals under the varying speed running condition,and the gearbox vibration signal in practical engineering are employed to validate the proposed method,respectively.As a result,the proposed method has the excellent performance to track the target ridge,and has a certain advantage compared with some existed ridge estimation methods.