Research On Crack Fault Vibration Response Modulation Mechanism And Remaining Useful Life Prediction Method Of Gear Transmission System

This thesis takes gear systems as the research object.Aiming at the problem of complex fault vibration response mechanism of gear systems,poor feature extraction ability as well as poor life prediction accuracy,etc.In-depth research on vibration response mechanism of gear systems crack fault,gear systems impact fault feature extraction method and remaining useful life prediction method have important theoretical research significance and engineering application value for gear systems health management.The main contents of the thesis are summarized as follows:(1)The rigid-flexible coupling dynamics model and the signal convolution model of fixed-axis gear train are established,the vibration response modulation mechanism of fixed-axis gear train under normal and crack faults are revealed.The excitation forces of the gearbox housing in fixed-axis gear train are the bearing support forces on the housing,and the flexible transmission paths are the frequency response functions between the bearing positions and the sensor point.Under the normal state of fixed-axis gear train,the frequency components of the vibration response signal are the meshing frequency and its harmonics.Affected by the frequency response function,the amplitude of each order meshing frequency no longer decreases with the increase of the frequency,but increases significantly near the natural frequency.Under the crack fault,in addition to the features under normal state,there are resonance modulation sidebands spaced by the fault gear rotation frequency.(2)Based on the signal convolution model of fixed-axis gear train,and comprehensively considering the influence of planetary gear train on various housing excitation forces and multiple flexible transmission paths,the rigid-flexible coupling dynamics model and the signal convolution model of planetary gear train are established to reveal the vibration response modulation mechanism of planetary gear train under normal,sun gear crack and ring gear crack faults.The signal convolution model of planetary gear train under the first kind of transmission path is established,which solves the problem that the traditional convolution theory is not suitable for the existence of time-varying excitation force and time-varying unit impulse response function.The excitation forces of the gearbox housing in planetary gear train are mainly the ring gear tooth excitation forces,the support forces of the sun gear bearing and the planet carrier bearing on the housing.The flexible transmission paths are the frequency response functions between the the ring gear tooth excitation forces position and the bearing position to the sensor point,respectively.Under the normal state of planetary gear train,the order spectrum characteristics of vibration response signal are centered on the meshing order and its higher orders which can be divisible by the planet gear passing order,and there are meshing modulation sidebands on both sides spaced by the planet gear passing order.Under the sun gear crack fault,in addition to the order features under normal state,there are three groups of modulation sidebands in the resonance region: the first group of resonance modulation sidebands are separated by N times the sun gear relative rotation order,the second and third groups of resonance modulation sidebands are separated by the planet gear passing order and the sun gear absolute rotation order.The second and third groups of modulation sidebands are distributed on both sides of the first group of modulation sidebands.Under the ring gear crack fault,in addition to the order features under normal state,there are only one group of resonance modulation sidebands spaced by N times ring gear relative rotation order.(3)According to the gear system signal convolution model and sparse decomposition theory,an impact sparse dictionary based on editing cepstrum to extract modal parameters is designed and a method of gear system impact fault feature extraction is proposed by using the ability of deconvolving the signal in cepstrum.The information of frequency response function is extracted by exponential window function in the cepstrum domain,and the multi-order modal parameters are obtained by rational fractional polynomial fitting,identification error of damping ratio caused by cepstrum windowing is compensated quantitatively;A sparse impact dictionary with unit impulse response function as atom is designed with clear physical meaning,and then the fault characteristic signal is reconstructed by the segmental matching pursuit algorithm inspired by the minimum fault period.The impact fault diagnosis of gear system is completed by demodulation analysis of reconstructed signals.The proposed method improves the identification accuracy of natural frequency and damping ratio.The sparse impact dictionary is not only similar to the fault impact waveform but also has clear physical meaning.Simulation and experimental analysis show that the proposed method has better anti-noise performance and diagnostic ability than the improved spectral kurtosis and traditional edited cepstrum methods.(4)Based on the researched vibration modulation mechanism of gear system and signal sparse decomposition,the sum of envelope squared amplitude spectrum of the reconstructed signal is constructed as the health index,and an improved particle filtering method for predicting the remaining useful life of gear system is proposed.The multi-dimensional feature distribution difference of original signal is calculated by Wasserstein distance,and the first prediction time of gear system is obtained adaptively.According to the sparse decomposition method,the fault degradation features are extracted,and then the envelope square amplitude spectrum is utilized to construct health index.The improved particle filter algorithm integrating parameter distribution adaptive adjustment method and parameter exponential smoothing method realizes the prediction of the remaining useful life of gear systems.The first prediction time identification method reduces the influence of manual setting of fault threshold.The constructed health index is directly associated with the fault mechanism features and has better anti-noise ability.The improved particle filter algorithm improves the adaptability and stability of remaining useful life prediction.Simulation and experimental analysis show that the proposed method has higher prediction accuracy and smaller fluctuation than the traditional particle filter algorithm.