Railside Acoustic Fault Diagnosis Technology Of Train Wheel Set Bearing Based On Doppler Distortion Signal

The wheelset rolling bearing of high speed railway is an indispensable component,which are prone to have damage under the train running.In case of early failure of the bearing,the damage will continue to expand without detected in time during the long running.Finally,the serious accidents are happened.Therefore,the evaluation and diagnosis of bearing under the safety state is necessary.Aiming at the condition monitoring and fault diagnosis of rolling bearings of trains,this thesis focus on the problems in the field of trackside acoustic technique such as spectrum distortion,extraction of weak information from mixed signals,effective noise reduction of target signals under strong noise background.Four non-stationary signal processing methods,such as(Local Mean Decomposition)LMD,(Fast Kutorgram)FK and spectral kurtosis theory,as well as optimized T-MUSIC(time-varying multiple signal classification)and AIR(angle interpolation resampling),are introduced into the fault diagnosis of train rolling bearings.The theoretical and application on the above-mentioned related problems is carried out.In order to solve the problem of acoustic spectrum aliasing,this thesis is based on microphone array signal correction,and proposes space correction method based on optimized time-varying multi-signal classification and angle interpolation resampling for multiple sound source signal.The method obtains the real-time location of the sound sources by optimized TMUSIC and establishes the resampling time series by relationship between the emission time and reception time of the sound source,so as to realized signal correction.Compared with the traditional method,this method has many advantages,such as no prior knowledge,less computation,strong noise robustness,suitable for variable speed sound source problems,etc.It has high potential in practical system applications.The trackside acoustic signal has weaker fault information,strong noise,and nonstationary feature after the correction of Doppler distortion.Based on this signal,improved compound interpolation envelope local mean decomposition(ICIE LMD)combined with fast kurtogram(FK)is proposed for train wheel-set rolling bearing.Local mean decomposition(LMD)is adaptive to decompose a signal into product function(PF)components.The accuracy of mono-components was influenced with the aim of constructing the local mean function and the envelope function.This thesis proposes ICIE LMD,which gives a simple way,to select the nonstationary coefficient that is applied to different parts of the signal for better accuracy and less computing overhead.FK is sensitive to impulse the signals,but it is easily influenced by non-Gaussian noise.First,the proposed method uses the ICIE LMD to decompose the signal into some PF components.The kurtosis index of PF is calculated to obtain the component that contains the most fault information.Then,FK is utilized to filter the selected one.The squared envelope spectrum is also used to obtain the fault frequency and identify the fault location.Simulation analysis is used to verify the benefits of ICIE LMD.The proposed method is utilized for the signal from the trackside microphone,which can diagnose the fault efficiently.