Research On Key Technology Of Rolling Bearing Fault Detection

Rolling bearings are the most important part of the rotating parts of the locomotive.The failure of the locomotive bearings will seriously affect the safe operation of the locomotive.In order to improve the feasibility of the rail-side acoustic rolling bearing fault detection system,this paper studies the key technologies of rolling bearing fault on-line detection which mainly includes the correction of Doppler distortion,the extraction of weak periodic signals from strong noise and the separation of multiple sound sources.In this paper,the main failure modes and natural vibration frequencies of NJP3226X1rolling bearing are analyzed firstly and the characteristic frequencies of four kinds of bearing parts?inner ring,outer ring,roller and cage?are further obtained.Then the modal analysis of train bearing parts is carried out by ANSYS,modeling is carried out in software according to measurement parameters,and finally the natural frequency values of each component are calculated which can be used as reference for fault detection.Then the Doppler distortion correction method is discussed.According to the sound radiation theory of moving sound sources,Doppler distortion signals can be obtained through simulation.Secondly,the process of distortion correction is analyzed:firstly,STFT is used to obtain the time-frequency distribution of the signal after downsampling,then Crazy Climber algorithm is used to estimate the instantaneous frequency of the signal and the least square method is used to fit to obtain the continuous value of the instantaneous frequency.Finally,the resampling time series is calculated and the signal is fitted with cubic spline function,the signal corrected for Doppler distortion is resampled according to the fitting result.In terms of strong noise,the bistable stochastic resonance weak signal detection method is studied which includes its scale transformation principle.Self-adaptive detection of weak periodic signals can be realized by adding signals or adjusting system parameters.The best matching stochastic resonance method is emphatically discussed here.Finally,the Woods-Saxon and Gaussian joint potential function is studied.the difference from SR system is that it has a flatter potential well and a steeper potential well wall and its five parameters are not coupled with each other.Through the relationship between the maximum output signal-to-noise ratio and the parameters,a better detection effect can be achieved.Finally,this paper uses time-frequency signal fusion and DFMS algorithm to achieve the purpose of sound source separation.The function of time-frequency signal fusion is to combine TFD with different window lengths through a data engine so as to obtain the improvement of signal resolution in time and frequency.DFMS algorithm mainly constructs a summation matrix.The value of each point in the summation matrix indicates that the matching template is consistent with the frequency distribution of the original signal under a fixed condition.The higher the point,the more matching it is.Fixed t₀,looking at the summation matrix from the f₀ direction,the peak value indicates the f₀location;Fixed f₀,observing the summation matrix from the t₀ direction and judging different time centers according to the peak value.Extracting the f₀ energy distribution at a specific time center can separate different sound sources.