| As one of the core components of vehicle running parts,the axle-box bearing has characteristic of complex structure,large load,and high running speed.Once the damage occurs,it will have a serious impact on the stability and safety of railway vehicles.It is of great significance to accurately identify the health status of axle-box bearings for the safe operation and scientific maintenance of railway vehicles.Due to the influence of the wheel-rail excitation,operating conditions and complex transmission paths,the fault characteristics of axle-box bearings are often obscured by complex background noise,and the traditional fault diagnosis methods of rotating machinery are difficult to be effective.Therefore,based on theoretical and experimental methods,the vibration response characteristics of the axle-box bearing system under wheel-rail excitation were analyzed,and combined with the impulsiveness and cyclo-stationarity characteristics of bearing fault signals,a axle-box bearing fault feature extraction method based on adaptive resonance demodulation was proposed.The main research contents of this paper are as follows:(1)The force and vibration characteristics of the axle-box bearing system are briefly analyzed.Then,the high-frequency excitation test of 50 ～ 300 km/h level is carried out on the single shaft rolling and vibrating test rig for the high-speed train.The vibration response of the axle-box bearing system under the excitation of track irregularity is collected,and the effects of speed level,bearing local defects and tread damage on the vibration response are analyzed.(2)The feature extraction method of weak fault of axle-box bearing under strong interference is studied.An improved kurtogram method with adjustable amplitude spectrum was proposed to enhance the fault feature components in vibration signals from from the perspectives of frequency domain noise reduction and time domain noise reduction.In the frequency domain,the signal is reconstructed by using the amplitude spectrum with different weight indexes,which is used as a preprocessing strategy to adjust the frequency component of the full frequency band.In time domain,the envelope of narrow-band filtered signal in the improved kurtogram is autocorrelated to further enhance the impact strength of bearing fault.The stability and effectiveness of the proposed method in the weak fault diagnosis of axle-box bearings are proved by simulation signals and several experimental signals.(3)The method of compound fault feature extraction of axle-box bearing system under strong disturbance is studied.In traditional kurtogram methods,binary wavelet packet decomposition and 1/3 binary tree structure are often used to generate kurtogram,but this method has the problem of insufficient frequency resolution.In addition,It is only suitable for single fault detection and cannot solve the problem of compound fault diagnosis.Aiming at the above shortcomings,an optimal resonance band recognition method based on the autocorrelation domain multi-point kurtogram(ACMKurtogram)is proposed.Firstly,this method can achieve adaptive segmentation of signal spectrum according to the minimum point distribution in the spectrum trend function,so as to obtain more accurate frequency band boundary matrix.Then,based on the rearrangement characteristics of periodic fault shocks in the autocorrelation domain,this method constructs a new index-the autocorrelation domain multi-point kurtosis(ACMK),which realizes the unified characterization of bearing fault impulsivenesss and cyclostationarity,and can effectively separate the compound fault features of axle-box bearing system.(4)The adaptive identification method of multi-resonance frequency bands of axle-box bearing system is studied.The calculation of ACMKurtogram needs to input the fault period as a priori knowledge.In view of the above shortcomings,a new Weight Kurtosis index is developed based on the impact and cyclo-stationarity of the fault signal.This index breaks through the heavy dependence on accurate fault cycle and avoids the influence of random shock harmonic components and white Gaussian noise.Based on Weight Kurtogram,and a new multi-resonance frequency band identification strategy is developed,which can highlight the fault impact intensity in each interval of the spectrum at the same time,and realize the separation and extraction of axle-box bearing compound faults or multiple faults of bearing and tread under wheel-rail excitation.(5)The fault feature extraction method of the axle-box bearing in variable speed mode is studied.In variable speed mode,the axle-box bearing fault signals no longer have cyclo-stationarity,and are also affected by strong interference components,which brings great difficulties to the identification of the optimal resonance band.In order to solve the above problems,this paper proposes a axle-box bearing fault feature extraction method based on iterative generalized demodulation.By iterative generalized demodulation of each narrowband filtered envelope signal,the stationary reset of the specific nonstationary axle-box bearing fault feature component and its frequency multiplier component is realized.Then,the kurtosis of the spectrum of generalized demodulation signals is calculated to evaluate the strength of bearing fault features in each narrowband signal,so as to locate the optimal demodulation frequency band.Simulation and experimental results show that the proposed method can effectively separate and extract fault characteristics of axle-box bearings under variable speed.An adaptive resonant band recognition method for axle-box bearing fault signals is studied.The fault characteristic information of axle-box bearing can be extracted from vibration signals of axle-box bearing system,so as to master the health status of axle-box bearing.The research results have certain theoretical value and practical engineering significance for ensuring safe and stable train operation and promoting the development of intelligent train operation and maintenance technology. |
PDF Full Text Request