High Resolution Time-Frequency Representation Method Of Synchroextracting Transform Theory And Its Application In Mechanical Fault Diagnosis

With the development of modern signal analysis theory,time-frequency(TF)analysis(TFA)technology has played an increasingly important role in mechanical fault diagnosis.It shoulders crucial functions in maintaining the normal operation of the equipment,and effectively avoiding the damage of key transmission parts and even the serious collapse of the whole facility.At present,large mechanical equipment generally presents the characteristics of complex internal structure and multi transmission bearing parts,which directly leads to the abnormal complexity of the vibration signal.At the same time,the complicated and varied service conditions and many external interference factors also make it difficult to effectively mine the fault information contained in vibration signals.In addition,the limitation of low time-frequency resolution of traditional time-frequency analysis methods further aggravates the difficulty of diagnosing the equipment failure.Under the above research backgrounds,this thesis is oriented to the following three typical problems in mechanical,namely the weak time-varying fault feature extraction at the early failure stage,the effective fault feature identification under strong background noise interference as well as the effective representation of adjacent time-varying feature,by deeply exploring the shortcomings of synchroextracting transform(SET)theory in mechanical fault diagnosis,a series of high resolution time-frequency representation methods are developed.Through the progressive step from theoretical exploration,simulation analysis to concrete application,the effectiveness of the proposed method in equipment health state operation and maintenance is verified.The research work of this thesis mainly includes the following aspects:(1)Making an in-depth study to the basic theory of SET,and aiming at the three typical problems,namely the weak time-varying fault feature extraction at the early failure stage,effective fault feature identification under strong background noise interference and effective representation of adjacent time-varying feature,the corresponding numerical simulation cases with the above fault characteristics are established to make a meticulous and thoughtful analysis to the shortcomings of the SET method in dealing with the above three typical problems,so as to lay a foundation for the improvement of SET in the subsequent chapters of this research.(2)Aiming at the problem of weak time-varying fault feature extraction at the early failure stage,a time-frequency analysis method called synchro spline-kernelled chirplet extracting transform(SSCET)was proposed.By utilizing the frequency rotating operator and the frequency shifting operator constructed by the spline kernel function,and uniting the idea of cycle iterative approximation,the changing law of the time-varying feature can be estimated effectively.Meanwhile,the idea of synchro extracting is adopted to further improve the energy concentration of the estimated time-varying feature.Under the synergistic effects of two frequency estimation operators,the idea of cyclic iterative approximation and synchro extracting,the weak time-varying fault feature at the early faulire stage can be extracted effectively.In addition,in view of the limitation of SSCET is not applicable to the analysis multi-component signal,a component separation strategy based on binary time-frequency image and connected component labeling method is studied,so as to provide an important guarantee for SSCET analyzing the vibration signals,and realizing the weak fault diagnosis at the early faulire stage of the equipment.(3)Aiming at the problem of effective fault feature identification under strong backgroung noise interference,a time-frequency analysis method called velocity synchronous chirplet extracting transform(VSCET)is proposed.The basis function with the synchronous variation laws of rotating speed built on the harmonic structure of rotating machinery vibration signal was introduced into the theoretical framework of SET,thus a velocity synchronus instantaneous frequency estimation operator for instantaneous frequency esitmation was constructed.On this basis,the synchro extracting operation of SET was further adopted to realize the robust estimation of time-varying feature information under strong background noise interference.In addition,by deeply exploring the inherent relationship between bearing fault characteristic coefficient,the rotating frequency information and instantaneous fault characteristic frequency,a fault identification technique based on VSCET and fault characteristic coefficient was studied to realize the effective identification of different bearing failures.(4)Aiming at the problem that the adjacent time-varying components contained in planetary gearbox and the wind turbine transmission chain are diffict to be represented effectively,a time-frequency method termed as matching synchroextracting transform(MSET)is proposed.By deeply investigaing the energy ambiguity of modulated parts in time-frequency spectrums,and utilizing the advantage of demodulated synchrosqueezing transform(DSST)in eliminating the energy ambiguity of modulated parts in time-frequency spectrum,the demodulation operator is introduced into SET theory to realize the accurate depict of feature information of time-frequency spectrum.In addition,aiming at shortcomings of the poor multi-component signal processing performance,as well as relying on the accurate prior instantaneous frequency frequencies of MSET,a demodulation filtering strategy-based multi-component separation thchnique,and a reliable initial instantaneous frequency estimation method based on the idea of order analysis and the second-order difference matrix are further studied.Based on the above series of contributions,the MSET method can generate a highly readable and high-resolution time-frequency representation result,such that the information of the adjacent time-varying characteristic components contained in planetary gearbox and wind turbine transmission chain can be well revealed.(5)The proposed high-resolution time-frequency representation techniques are applied to the fault diagnosis of some specific objects such as the rolling bearing,the planetary gearbox and the wind turbine transmission chain,so as to verify the effectiveness of the proposed techniques in mechanical fault diagnosis.